List your company

Style Manipulation Attack on Qwen3-8B

2ScoreJudgeDeceiver
-0.080.4611.54May 24, 2026

Evaluation results

MethodLinks
JudgeDeceiverOptimization Steps=600, Attack Type=White-box2026.052
BITEOptimization Steps=25, Attack Type=Black-box2026.051.905
JudgeDeceiverOptimization Steps=300, Attack Type=White-box2026.051.588
JudgeDeceiverOptimization Steps=200, Attack Type=White-box2026.051.438
JudgeDeceiverOptimization Steps=100, Attack Type=White-box2026.050.889
JudgeDeceiverOptimization Steps=25, Attack Type=White-box2026.050